Title :
Scalable sparse approximation of a sample mean
Author :
Cortes, Efren Cruz ; Scott, Clayton
Author_Institution :
Electr. Eng. & Comput. Sci., Univ. of Michigan, Ann Arbor, MI, USA
Abstract :
We examine the problem of approximating the mean of a set of vectors as a sparse linear combination of those vectors. This problem is motivated by a common methodology in machine learning where a probability distribution is represented as the sample mean of kernel functions. In applications where this kernel mean function is evaluated repeatedly, having a sparse approximation is essential for scalability. However, existing sparse approximation algorithms such as matching and basis pursuit scale quadratically in the sample size, and are therefore not well suited to this problem for large sample sizes. We introduce an approximation bound involving a novel incoherence measure, and propose bound minimization as a sparse approximation strategy. In the context of sparsely approximating a kernel mean function, the bound is efficiently minimized by solving an appropriate instance of the k-center problem, and the resulting algorithm has linear complexity in the sample size.
Keywords :
approximation theory; computational complexity; learning (artificial intelligence); minimisation; sampling methods; statistical distributions; vectors; bound minimization; k-center problem; kernel mean functions; linear complexity; machine learning; mean approximation; probability distribution; pursuit matching; sample mean; scalable sparse approximation algorithm; vectors set; Approximation algorithms; Approximation methods; Context; Estimation; Kernel; Signal processing algorithms; Vectors; incoherence; k-center problem; kernel methods; sparse approximation;
Conference_Titel :
Acoustics, Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on
Conference_Location :
Florence
DOI :
10.1109/ICASSP.2014.6854602